Spaces:
Runtime error
Runtime error
File size: 3,578 Bytes
d59aeff e42c164 d59aeff e42c164 d59aeff 498fbde e42c164 d59aeff a0907be |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 |
import gradio as gr
import re
import random
import string
import librosa
import numpy as np
from pathlib import Path
from scipy.io.wavfile import write
from encoder import inference as encoder
from vocoder.hifigan import inference as gan_vocoder
from synthesizer.inference import Synthesizer
class Mandarin:
def __init__(self):
self.encoder_path = "encoder/saved_models/pretrained.pt"
self.vocoder_path = "vocoder/saved_models/pretrained/g_hifigan.pt"
self.config_fpath = "vocoder/hifigan/config_16k_.json"
self.accent = "synthesizer/saved_models/普通话.pt"
synthesizers_cache = {}
if synthesizers_cache.get(self.accent) is None:
self.current_synt = Synthesizer(Path(self.accent))
synthesizers_cache[self.accent] = self.current_synt
else:
self.current_synt = synthesizers_cache[self.accent]
encoder.load_model(Path(self.encoder_path))
gan_vocoder.load_model(Path(self.vocoder_path), self.config_fpath)
def setVoice(self, timbre):
self.timbre = timbre
wav, sample_rate, = librosa.load(self.timbre)
encoder_wav = encoder.preprocess_wav(wav, sample_rate)
self.embed, _, _ = encoder.embed_utterance(encoder_wav, return_partials=True)
def say(self, text):
texts = filter(None, text.split("\n"))
punctuation = "！，。、？!,.?：:" # punctuate and split/clean text
processed_texts = []
for text in texts:
for processed_text in re.sub(r'[{}]+'.format(punctuation), '\n', text).split('\n'):
if processed_text:
processed_texts.append(processed_text.strip())
texts = processed_texts
embeds = [self.embed] * len(texts)
specs = self.current_synt.synthesize_spectrograms(texts, embeds)
spec = np.concatenate(specs, axis=1)
wav, sample_rate = gan_vocoder.infer_waveform(spec)
return wav, sample_rate
def greet(audio, text, voice=None):
print(f"Log print: audio name=[{audio.name}], text=[{text}]")
if voice is None:
voice = Mandarin()
voice.setVoice(audio.name)
voice.say("加载成功")
wav, sample_rate = voice.say(text)
output_file = "".join( random.sample(string.ascii_lowercase + string.digits, 11) ) + ".wav"
write(output_file, sample_rate, wav.astype(np.float32))
return output_file, voice
def main():
demo = gr.Interface(
fn=greet,
inputs=[gr.inputs.Audio(type="file"),"text", "state"],
outputs=[gr.outputs.Audio(type="file"), "state"],
title="Tacotron Zero-short Voice Clone (Chinese Version)"
)
demo.launch()
def new_main():
with gr.Blocks() as demo:
statas = gr.State()
title = gr.Markdown("# Tacotron Zero-short Voice Clone (Chinese Version)")
with gr.Row():
with gr.Column(scale=1):
input_audio = gr.Audio(type="file", label="Source Audio", value="exp/lihao_01.wav")
input_text = gr.Textbox(value="大家好，我是正在搬砖的李昊，这是一段合成音频。")
with gr.Row():
clear = gr.ClearButton()
submit = gr.Button(value="Submit", variant='primary')
with gr.Column(scale=1):
output_audio = gr.Audio(type="file", label="Output Audio")
_ = submit.click(greet, inputs=[input_audio, input_text, statas], outputs=[output_audio, statas])
demo.launch()
if __name__=="__main__":
new_main()
|